Obstacle detection assitance system and obstacle detection assistance method
The obstacle detection support system in track transportation systems addresses the issue of false alarms by using adaptive detection criteria based on environmental conditions, ensuring reduced false alarms and uninterrupted driver support, thus improving safety and operability.
Patent Information
- Application Number
- PCT/JP2024/015167
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-10
- Filing Date
- 2024-04-16
- Publication Date
- 2024-11-14
AI Technical Summary
Existing obstacle detection systems in track transportation systems face challenges in reducing false alarms without interrupting driving assistance, as improved detection rates often lead to increased false positive rates, causing unnecessary interruptions in driver support.
An obstacle detection support system that includes a driving support display unit, an obstacle detection unit, and a determination criterion determination unit, which sets thresholds for the number of times of object detection and the degree of coincidence of detection results based on the external environment, allowing for adaptive criteria to minimize false alarms without disrupting driver support.
The system effectively reduces false alarms in track transportation systems while maintaining continuous driving support by dynamically adjusting detection criteria according to environmental conditions, thereby enhancing safety and operability.
Smart Images

Figure JP2024015167_14112024_PF_FP_ABST
Abstract
Description
Obstacle detection support system and obstacle detection support method
[0001] The present invention relates to an obstacle detection support system mounted on a rail transportation system that travels on a fixed rail.
[0002] In a track transportation system in which transportation vehicles travel on a track, if an obstacle is present on the track, it is not possible to avoid it by steering. Therefore, detecting obstacles on the track is important for improving the safety and operability of the track transportation system. In recent years, research has been conducted on obstacle detection support systems that detect obstacles on the track using external sensors such as millimeter-wave radar, laser radar, and cameras, and support the driver's driving by notifying the driver of the detection results. In an obstacle detection support system, it is important to reduce false detection of obstacles in order to avoid cumbersome notifications to the driver. Patent Document 1 discloses a technology for reducing false alarms to the driver in an obstacle detection support system.
[0003] Japanese Patent Application Laid-Open No. 2022-135627
[0004] In an obstacle detection assistance system, it is important to reduce false detection of obstacles to avoid cumbersome notifications to the driver. Generally, improving the detection rate tends to increase the false detection rate as well. The technology described in Patent Document 1 suspends driving assistance to the driver when the train's external environment meets a false detection condition in which the obstacle detection unit erroneously detects a non-obstacle as an obstacle. Therefore, while false detections to the driver are reduced, there is a problem in that driving assistance is not provided in sections where the false detection condition applies.
[0005] In order to address the above-mentioned problems, the present invention aims to reduce false alarms in an obstacle detection assistance system installed in a rail transportation system without interrupting driving assistance to the driver.
[0006] In order to solve the above problems, the obstacle detection assistance system according to the present invention is a system that assists in obstacle detection during vehicle driving, and comprises a driving assistance display unit that displays driving assistance information that assists vehicle driving, an obstacle detection unit that detects obstacles that hinder the driving of the vehicle, a judgment criterion determination unit that determines the judgment criterion used by the obstacle detection unit to detect an obstacle, and an external environment recognition unit that recognizes the external environment of the vehicle, wherein the judgment criterion determination unit sets a judgment criterion according to the external environment, and the judgment criterion is set to be a threshold value for the number of times an obstacle is detected by the obstacle detection unit, or a degree of agreement of the detection results by the obstacle detection unit, or both the threshold value for the number of times an obstacle is detected and the degree of agreement of the detection results.
[0007] According to the present invention, in an obstacle detection assistance system installed in a rail transportation system, it is possible to reduce false alarms without interrupting driving assistance to the driver.
[0008] Other objects, configurations and effects will become apparent from the detailed description of the invention and the drawings.
[0009] FIG. 1 is a diagram showing the configuration of an obstacle detection support system in a first embodiment of the present invention. FIG. 2 is a flowchart showing the processing of a determination criterion determination unit in the first embodiment of the present invention. FIG. 3 is a functional block diagram showing the processing of an obstacle detection unit in the first embodiment of the present invention. FIG. 4 is a diagram showing obstacle determination for each threshold value executed by an obstacle determination unit in the first embodiment of the present invention. FIG. 5 is a flowchart showing the processing of a determination criterion determination unit in a second embodiment of the present invention. FIG. 6 is a diagram showing obstacle determination for each detection method executed by an obstacle determination unit in the second embodiment of the present invention.
[0010] 1 is a diagram showing the configuration of an obstacle detection assistance system. The present invention comprises an external sensor 110 mounted on a train 100 that senses the external environment, and an obstacle detection assistance system 120 that notifies the driver of the presence or absence of obstacles ahead and assists the driver in driving. The obstacle detection assistance system 120 comprises an external environment recognition unit 121 that recognizes the external environment of the train 100 based on sensor information from the external sensor 110, a judgment criterion determination unit 122 that determines judgment criteria for obstacle detection based on the information from the external environment recognition unit 121, an obstacle detection unit 123 that detects obstacles ahead of the train 100 based on the sensor information from the external sensor 110 and the judgment criteria from the judgment criterion determination unit 122, and a driving assistance display unit 124 that notifies the driver of assistance content information based on the obstacle detection result information from the obstacle detection unit 123.
[0011] The external sensor 110 senses the conditions around the train 100 (particularly ahead) and transmits the sensing data to the external environment recognition unit 121 and the obstacle detection unit 123. The external sensor 110 includes a camera, a LIDAR (laser range finder), a millimeter-wave radar, a GNSS (Global Navigation Satellite System), etc. The camera includes a monocular camera, a stereo camera, an infrared camera, etc. Generally, multiple sensors are installed for redundancy. When the train 100 moves forward in the direction of travel, the external sensor 110 installed in the leading car on the side facing the direction of travel is used.
[0012] The external environment recognition unit 121 recognizes the external environment ahead of the train 100 using external sensor data from the external sensor 110. The recognized external environment includes the route environment, such as stations, railroad crossings, and railroad bridges; the route shape environment, such as straight lines and curves; the time environment, such as daytime and nighttime; and the train environment, such as vehicle speed and vehicle position. The route environment is primarily recognized using a deep neural network (DNN) trained to learn the target. The route shape environment is recognized from the shape of the rail ahead detected by image processing. Specific methods for detecting rails using image processing include searching for rails using brightness gradient information in an image and detecting rails using semantic segmentation using a DNN. In the present invention, the method is not important as long as the rails can be detected. The time environment may be estimated by estimating the time zone from the average brightness value of external sensor information (especially image data), or by using time information included in GNSS data. Time information may also be acquired from another on-board system (not shown). In the present invention, the method is not important as long as the time can be estimated. Regarding the train environment, the vehicle speed may be estimated from GNSS data, or the speed obtained by performing SLAM (Simultaneous Localization and Mapping) using LIDAR data may be used. The speed may also be acquired from another on-board system (not shown). The train's own position may be latitude and longitude information obtained from GNSS data, or position information based on a certain position obtained by integrating the train speed. The train's own position may also be acquired from another on-board system (not shown). In the present invention, any method is acceptable as long as the train speed and the own position can be estimated. The external environment recognition unit 121 transmits the recognized external environment to the judgment criterion determination unit 122. A database may be maintained that records the relationship between environmental information about the route environment and the route shape environment and the location where the environment exists, and the external environment information about the current location of the train 100 may be obtained from the database based on the position of the train 100.
[0013] The judgment criterion determination unit 122 determines the judgment criterion for object detection by the obstacle detection unit 123 using the external environment recognized by the external environment recognition unit 121, vehicle information acquired from other on-board systems (not shown), the vehicle's own position, and door opening / closing information. The judgment criterion determination unit 122 determines whether the external environment is an external environment that may affect false detection of objects, and determines the judgment criterion for object detection according to the external environment. The judgment criterion includes continuous detection threshold information, which indicates how many consecutive detections are required in the time-series data of the object detection results to determine a detection when the obstacle detection unit 123 detects an object, and detection method information, which indicates whether the object detection results are determined as a detection by logical sum or logical product. The judgment criterion determined by the judgment criterion determination unit 122 is the continuous detection threshold information, the detection method information, or both the continuous detection threshold information and the detection method information. The judgment criterion determination unit 122 transmits the determined judgment criterion to the obstacle detection unit 123.
[0014] The obstacle detection unit 123 grasps the situation ahead of the train 100 and detects obstacles using external sensor data from the external sensor 110 and the criteria from the criteria determination unit 122. The obstacle detection unit 123 can use technology used in the automotive field for its object detection process. For example, one method is to create parallax images using a stereo camera and recognize the shape and position of an object ahead from the parallax images. Other methods include recognizing objects in monocular images using a DNN, or recognizing objects from LIDAR point cloud data. Here, a DNN is one of the methods used in machine learning, and detects various objects by extracting and learning the features of the objects. In the present invention, the method is not important as long as it can detect obstacles. The obstacle detection unit 123 makes a final obstacle determination based on the object detection results from the stereo camera, machine learning, and LIDAR in accordance with the criteria from the criteria determination unit 122. For example, if "2" is sent as the consecutive detection threshold information from the judgment criterion determination unit 122, an object is not considered to have been detected unless it is detected two consecutive times. This reduces the possibility of misidentifying randomly detected objects as obstacles, thereby reducing false detections.
[0015] The driving assistance display unit 124 determines the assistance content to be notified to the driver based on the obstacle detection result information from the obstacle detection unit 123 and the vehicle speed, and displays it on a display device. For example, a scene is defined for each environment ahead of the train 100, and the assistance content to the driver is determined for each scene. Examples of scenes include: (1) There are no obstacles on or around the track. (2) There is an obstacle around the track. (3) There is an obstacle on the track, but it is far from the train 100 or the speed of the train 100 is low, so there is no need to apply the brakes to avoid a collision. (4) There is an obstacle on the track, but the collision can be avoided by immediately applying the brakes. (5) There is an obstacle on the track, and there is a high possibility that a collision cannot be avoided.
[0016] The driving assistance display unit 124 determines the assistance content for each scene, for example, if the scene ahead of the train 100 is (3), the driver is notified only of the presence of an object, and if it is (4), the driver is notified to apply the brakes.
[0017] The driving assistance display unit 124 notifies the driver of the assistance provided via an HMI (Human Machine Interface) or by voice. In the present invention, the method of transmission is not important as long as the assistance provided to the driver is transmitted to the driver.
[0018] Next, the processing of the determination criterion determination unit 122 according to the first embodiment of the present invention will be described with reference to Fig. 2. In the first embodiment, an example will be described in which the determination criterion output by the determination criterion determination unit 122 is continuous detection threshold information.
[0019] FIG. 2 is a flowchart showing the process executed by the criterion determination unit 122 .
[0020] Step S201: The judgment criteria determination unit 122 acquires external environment information (route environment such as stations, railroad crossings, and railway bridges, route shape environment such as straight lines and curves, time environment such as day or night, and train environment such as vehicle speed) from the external environment recognition unit 121.
[0021] Step S202: The criterion determination unit 122 acquires the vehicle speed from another on-board system (for example, a vehicle information control system). The vehicle speed may be the vehicle speed estimated by the external environment recognition unit 121. In the present invention, it is sufficient to obtain the vehicle speed, and the calculation method or acquisition method is not important.
[0022] Step S203: The criterion determination unit 122 acquires the vehicle's own position from another on-board system (for example, a vehicle information control system). The vehicle's own position may be the one estimated by the external environment recognition unit 121. In the present invention, it is sufficient to obtain the vehicle's own position, and the calculation method and acquisition method are not important.
[0023] Step S204: The determination criterion determination unit 122 acquires door open / close information indicating the open / close state of the doors from another on-board system (for example, a vehicle information control system). In the present invention, it is sufficient to acquire the door open / close information, and the calculation method and acquisition method thereof are not important.
[0024] Step S205: The judgment criteria determination unit 122 determines whether the judgment criteria should be changed based on the external environment information. If the route environment is not a station, railroad crossing, or railway bridge, the possibility of false detection is low, so it is determined that the judgment criteria do not need to be changed. Proceed to step S208. If the route environment is any of a station, railroad crossing, and railway bridge, the possibility of false detection is high, so it is determined that the judgment criteria need to be changed. Proceed to step S206. If the route shape environment is straight, the possibility of false detection is low, so it is determined that the judgment criteria do not need to be changed. Proceed to step S208. If the route shape environment is curved, the possibility of false detection is high, so it is determined that the judgment criteria need to be changed. Proceed to step S206.
[0025] Step S206: The judgment criterion determination unit 122 judges whether the train 100 is moving or not based on the vehicle speed. If the train 100 is moving, it is determined that the judgment criterion needs to be changed because there is a high possibility of false detection. The process proceeds to step S209. If the train is stopped, the process proceeds to step S207.
[0026] Step S207: The judgment criteria determination unit 122 judges whether the doors of the train 100 are open or closed from the door opening / closing information. If the doors are closed, it is determined that the train 100 is about to depart, and the judgment criteria are changed so that obstacles near the train 100 can be detected when the train departs. Proceed to step S210. If the doors are open, it is determined that the train is stopping at the station. If an alert is given to the driver every time an obstacle is detected ahead of the train 100 while the train is stopping at the station, the support display may become cumbersome. Therefore, the judgment criteria are changed so that obstacles are not generally detected while the train is stopping at the station. Proceed to step S211.
[0027] Step S208: This is a case where the external environment of the train 100 is determined to be an environment where the possibility of false detection is low. Therefore, in order to reduce non-detection, a continuous detection threshold is set so that an object is detected as much as possible. For example, if an object is detected twice in a row, it is considered a detection. The threshold may be other than "2". It is desirable to perform an evaluation using external sensor data acquired in the environment where the obstacle detection assistance system 120 is installed, and to set a threshold where non-detection and false detection are within a desired range. Therefore, if false detection is within the desired range, the continuous detection threshold may be set to "1" (i.e., if an object is detected, it is considered a detection). The continuous detection threshold determined in this step is used as the first threshold. Proceed to step S212.
[0028] Step S209: This is a case where the external environment of the train 100 is determined to be an environment in which there is a relatively high possibility of false detection. Therefore, in order to reduce false detection, a continuous detection threshold is set so that an object is not detected as much as possible. The threshold defined in this step is a value larger than the first threshold, and for example, a detection is determined if an object is detected four times in a row. The threshold may be a value other than "4", and it is desirable to evaluate using external sensor data acquired in the environment in which the obstacle detection assistance system 120 is installed, and to set a threshold such that false detection falls within a desired range. The continuous detection threshold determined in this step is set as the second threshold. Proceed to step S212.
[0029] Step S210: This is a case where the external environment of the train 100 is determined to be stopped and about to depart. In this case, it is necessary to detect obstacles near the train 100 when it departs, but because the train is stopped, there is a possibility of repeatedly falsely detecting the same object. Therefore, the threshold defined in this step is a value greater than the second threshold to make false detection of objects less likely than when the train is moving. However, as described above, since it is necessary to detect obstacles near the train 100, it is desirable to set the difference from the second threshold to a small value. For example, five consecutive detections are deemed to be a detection. The threshold may be other than "5." It is desirable to evaluate using external sensor data acquired in the environment where the obstacle detection assistance system 120 is installed and determine a threshold that falls within a desired range of false detections. The continuous detection threshold determined in this step is designated as the third threshold. Proceed to step S212.
[0030] Step S211: This is a case where the external environment of the train 100 is determined to be a state where the train is stopped at a station and the doors are open. In this case, if the driver is notified every time an obstacle is detected ahead of the train 100, the assistance display may become cumbersome. Furthermore, since the train 100 is stopped, even if no obstacle is detected, there is no possibility of a collision with the obstacle. Therefore, a large value is set as the threshold in this step so that objects are basically not detected. For example, a value equal to or greater than the maximum number of time-series data that the obstacle detection unit 123 can hold for use in obstacle detection is set as the threshold. In this way, objects ahead of the train 100 are not detected, and assistance displays to the driver when the train is stopped at a station and the doors are open can be minimized. The continuous detection threshold determined in this step is set as the fourth threshold. Proceed to step S212.
[0031] Step S212: The determination criterion determination unit 122 transmits the determination criterion to the obstacle detection unit 123.
[0032] Next, the processing of the obstacle detection unit 123 according to the first embodiment of the present invention will be described with reference to FIG.
[0033] FIG. 3 is a functional block diagram showing the processing executed by the obstacle detection unit 123.
[0034] The sensor data acquisition unit 301 of the obstacle detection unit 123 acquires sensor data from an external sensor 110 such as a stereo camera, a monocular camera, LIDAR, or GNSS. The external sensor 110 may be any type of external sensor 110 capable of recognizing the external environment. The sensor data acquisition unit 301 transmits image data to a high-visibility processing unit 302, LIDAR data to an object detection unit (LIDAR) 306, and GNSS data to a vehicle speed calculation unit 307. The high-visibility processing unit 302 adjusts the brightness of the image data and corrects scenes with differences in brightness, such as tunnel entrances and exits, to generate a highly visible image. The generated high-visibility image is transmitted to a traveling area detection unit 303, an object detection unit (stereo) 304, and an object detection unit (machine learning) 305. The traveling area detection unit 303 detects rails through image processing and defines a traveling area in which the train 100 may travel based on the rail detection results. The travel area may be defined based on a method other than rail detection results. For example, a database associating rail shapes with positions may be maintained, and the rail shape ahead of the train 100 may be determined by comparing the train's current position with the position in the database, and the travel area may be defined based on the determined shape. Furthermore, if the external environment recognition unit 121 performs rail detection processing, rail detection results may be acquired from the external environment recognition unit 121. In the present invention, any method is acceptable as long as the travel area can be defined. The object detection unit (stereo) 304 creates parallax images using a stereo camera and recognizes the three-dimensional shape and position of objects ahead from the parallax images. The detection results are transmitted to the obstacle determination unit 308. The object detection unit (machine learning) 305 recognizes objects in the image using a DNN from the monocular camera image. A DNN is one of the methods used in machine learning, and detects various objects by extracting and learning the features of the objects. The object detection unit (machine learning) 305 transmits the detection results to the obstacle determination unit 308. The object detection unit (LIDAR) 306 detects objects from point cloud data acquired from the LIDAR sensor. Point clouds that exist in close proximity are considered to be a single object, and the three-dimensional shape and position of objects around the train 100 are recognized. The object detection unit (LIDAR) 306 transmits the detection results to the obstacle determination unit 308.The vehicle speed calculation unit 307 calculates the vehicle speed from the position information (latitude and longitude) obtained from the GNSS data and the elapsed time, and transmits it to the obstacle determination unit 308. The obstacle determination unit 308 determines an obstacle to be notified to the driver based on the driving area acquired from the driving area detection unit 303, the object detection results acquired from the object detection unit (stereo) 304, the object detection unit (machine learning) 305, and the object detection unit (LIDAR) 306, the vehicle speed acquired from the vehicle speed calculation unit 307, and the judgment criterion acquired from the judgment criterion determination unit 122.
[0035] Next, an example of the obstacle determination process for each threshold value executed by the obstacle determination unit 308 in FIG. 3 will be described with reference to FIG. 4. FIG. 4 shows the time series results (n, n-1, n-2, ...) of a certain detection result and the output for each threshold value when the detection result is input. In the example of FIG. 4, the first threshold is "2," the second threshold is "4," the third threshold is "5," and the fourth threshold is "6 or more." In FIG. 4, an input "O" indicates that a detected object was detected, and an "X" indicates that no detected object was detected. An output "O" indicates that the output of the obstacle detection unit 123 has determined that an obstacle is present, and an "X" indicates that no obstacle is present. The object detection results input here are the detection results of objects present within the running area and within the stopping distance of the train 100 calculated from the vehicle speed, among the object detection results obtained from the object detection unit (stereo) 304, object detection unit (machine learning) 305, and object detection unit (LIDAR) 306 in FIG. 3. In the example shown in FIG. 4 , if an object is detected only twice consecutively among the time-series inputs, the judgment criterion determining unit 122 determines that an obstacle is present if the judgment criterion is the first threshold, but determines that no obstacle is present if the criterion is other than the first threshold. Therefore, by changing the judgment criterion (threshold), the judgment criterion determining unit 122 can output different obstacle detection results for the same detection result input. If the judgment criterion is the second threshold, if no object is detected in the last four consecutive time-series inputs, it is determined that no obstacle is present. Therefore, even if false detections occur consecutively, notifications to the driver are suppressed, making it possible to reduce false alarms without interrupting driving assistance. If the judgment criterion is the fourth threshold, a determination that no obstacle is present is made regardless of the input result. Therefore, it is possible to limit notifications to the driver when the train is stopped at a station and the doors are open. This makes it possible, for example, to prevent pedestrians crossing the street from being identified as obstacles and receiving assistance alerts to the driver at a station with the train doors open, at a station where a passenger crossing is located immediately before the train 100.
[0036] In this embodiment, the recognized external environment includes the track environment, such as stations, railroad crossings, and railroad bridges; the track shape environment, such as straight lines and curves; the time environment, such as daytime and nighttime; and the train environment, such as vehicle speed and vehicle position. However, other environments that are prone to false detection may also be included as recognition targets. Examples include weather (rain, snow), backlighting, and passing an oncoming train at night. These conditions are particularly challenging for image recognition and increase the likelihood of false detection. Weather information may be acquired from an external system (not shown), or rain may be determined if the wipers are operating. Backlighting and passing an oncoming train at night may be determined based on the average brightness value of the entire image, or by detecting an area with extremely high brightness values. In the present invention, the method for recognizing external conditions (weather (rain, snow), backlighting, and passing an oncoming train at night) is not critical, as long as the external environment recognition unit 121 can recognize them.
[0037] In this embodiment, the judgment criteria are determined based on the external environment recognized by the external environment recognition unit 121. However, if the route environment, route shape environment, time environment, and train environment can be obtained from a source other than the external environment recognition unit 121, the route environment, route shape environment, time environment, and train environment may be obtained from a source other than the external environment recognition unit 121. For example, a database in which the route environment and route shape environment are associated with position information may be maintained, and the route environment and route shape environment may be obtained by comparing the train 100's own position with the position in the database. Furthermore, since the time environment and train environment are generally held by other on-board systems (e.g., a train information control system), the time environment and train environment may be obtained from the other on-board systems. In this way, it is possible to reduce the processing load on the external environment recognition unit 121.
[0038] 5 is a flowchart showing the processing executed by the judgment criterion determination unit 122 in the second embodiment of the obstacle detection assistance system 120. In the first embodiment, a continuous detection threshold was used as the judgment criterion, and the processing of the obstacle determination unit 308 was controlled by changing the continuous detection threshold to reduce false detections, but in this embodiment, an example will be described in which the detection method of determining whether the object detection results are determined by logical sum or logical product is used as the judgment criterion, and the processing of the obstacle determination unit 308 is controlled by changing the detection method to reduce false detections.
[0039] The processing of the determination criterion determination unit 122 according to the second embodiment of the present invention will be described with reference to Fig. 5. Fig. 5 is a flowchart showing the processing executed by the determination criterion determination unit 122.
[0040] Step S501: The judgment criteria determination unit 122 acquires external environment information (route environment such as stations, railroad crossings, and railway bridges, route shape environment such as straight lines and curves, time environment such as day or night, and train environment such as vehicle speed) from the external environment recognition unit 121.
[0041] Step S502: The criterion determination unit 122 acquires the vehicle speed from another on-board system (for example, a vehicle information control system). The vehicle speed may be the vehicle speed estimated by the external environment recognition unit 121. In the present invention, it is sufficient to obtain the vehicle speed, and the calculation method or acquisition method is not important.
[0042] Step S503: The criterion determination unit 122 acquires the vehicle's own position from another on-board system (for example, a vehicle information control system). The vehicle's own position may be the one estimated by the external environment recognition unit 121. In the present invention, it is sufficient to obtain the vehicle's own position, and the calculation method and acquisition method are not important.
[0043] Step S504: The determination criterion determination unit 122 acquires door open / close information indicating the open / close state of the doors from another on-board system (for example, a vehicle information control system). In the present invention, it is sufficient to acquire the door open / close information, and the calculation method and acquisition method thereof are not important.
[0044] Step S505: The judgment criteria determination unit 122 determines whether the judgment criteria should be changed based on the external environment information. If the route environment is not a station, railroad crossing, or railway bridge, the possibility of erroneous detection is low, so it is determined that the judgment criteria do not need to be changed. Proceed to step S507. If the route environment is any of a station, railroad crossing, and railway bridge, the possibility of erroneous detection is high, so it is determined that the judgment criteria need to be changed. Proceed to step S506. If the route shape environment is straight, the possibility of erroneous detection is low, so it is determined that the judgment criteria do not need to be changed. Proceed to step S507. If the route shape environment is curved, the possibility of erroneous detection is high, so it is determined that the judgment criteria need to be changed. Proceed to step S506.
[0045] Step S506: The judgment criteria determination unit 122 determines whether the train 100 is moving or whether the doors are closed based on the vehicle speed and door opening / closing information. If the train is moving, there is a high possibility of false detection, so it is determined that the judgment criteria need to be changed. Also, if the train is not moving but the doors are closed, it is determined that the train 100 is about to depart, and the judgment criteria are changed so that obstacles near the train 100 can be detected at the time of departure. Proceed to step S508. If the train is stopped and the doors are open, it is determined that the train is stopped at the station. If the driver is notified every time an obstacle is detected ahead of the train 100 while the train is stopped at the station, the support display may become cumbersome. Therefore, the judgment criteria are changed so that obstacles are not generally detected while the train is stopped at the station. Proceed to step S509.
[0046] Step S507: This is a case where the external environment of the train 100 is judged to be an environment where the possibility of false detection is low. Therefore, in order to reduce the number of undetected objects, a judgment criterion is set to detect objects as much as possible. For example, when judging an obstacle using three time-series results of obstacle detection, the logical sum of the three time-series results is set as the output of the obstacle detection. In other words, if an object is detected at least once out of the three time-series results, it is judged to have been detected. In this way, it is possible to reduce the number of undetected objects. The "logical sum" is set as the judgment criterion as the detection method. Proceed to step S510.
[0047] Step S508: This is a case where the external environment of the train 100 is judged to be an environment where there is a relatively high possibility of false detection. Therefore, in order to reduce false detection, a judgment criterion is set that will detect objects as little as possible. For example, when judging an obstacle using three time series results of obstacle detection, the logical product of the three time series results is set as the output of obstacle detection. In other words, if an object is detected three times out of the three time series results, it is considered to have been detected. In this way, it is possible to eliminate false detection of objects that occur randomly, and reduce false detections in the driving assistance that is notified to the driver. The "logical product" is set as the judgment criterion for the detection method. Proceed to step S510.
[0048] Step S509: This is a case where the external environment of the train 100 is determined to be a state where the train is stopped at a station and the doors are open. In this case, if an alert is given to the driver every time an obstacle is detected ahead of the train 100, the assistance display may become cumbersome. Furthermore, since the train 100 is stopped, even if no obstacle is detected, there is no possibility of a collision with the obstacle. Therefore, a determination criterion is set so that no object is detected. For example, if an obstacle is determined using three time-series results of obstacle detection, a "no obstacle" message is output regardless of the three time-series results. In this way, no objects ahead of the train 100 are detected, and assistance displays to the driver when the train is stopped at a station and the doors are open can be minimized. The detection method is determined to be "no obstacle regardless of input results." Proceed to step S510.
[0049] Step S510: The determination criterion determination unit 122 transmits the determination criterion to the obstacle detection unit 123.
[0050] Next, an example of the obstacle determination process for each determination criterion executed by the obstacle determination unit 308 in FIG. 3 in the second embodiment will be described with reference to FIG. 6 . FIG. 6 shows a time series of detection results (n, n-1, n-2, ...) and the output for each determination criterion when the detection results are input. In the example of FIG. 6 , three types of obstacle detection methods are used as the determination criterion: "logical sum of inputs," "logical product of inputs," and "no obstacle regardless of input." In FIG. 6 , an input "o" indicates that a detected object was detected, and an "x" indicates that no detected object was detected. An output "o" indicates that the output of the obstacle detection unit 123 has determined that an obstacle is present, and an "x" indicates that no obstacle is present. The object detection results input here are the detection results of objects present within the travel area and within the stopping distance of the train 100 calculated from the vehicle speed, among the object detection results obtained from the object detection unit (stereo) 304, object detection unit (machine learning) 305, and object detection unit (LIDAR) 306 in FIG. 3 . When the judgment criterion is defined as "logical sum," an obstacle is determined to be present if it is detected in at least one of the three time-series results. This reduces the rate of non-detection. On the other hand, when the external environment is determined to be one in which false detection is likely to occur, the detection method of the judgment criterion is set to logical sum. When the judgment criterion is "logical sum," an obstacle is determined to be present if it is detected in all three time-series results. Therefore, even if false detection occurs randomly, if it does not occur three times in a row, a notification to the driver is suppressed, making it possible to reduce false alarms without interrupting driving assistance. When the judgment criterion is "no obstacle regardless of input," an obstacle is determined to be absent regardless of the input result, making it possible to limit notifications to the driver when the train is stopped at a station and the doors are open. This makes it possible, for example, to prevent pedestrians crossing the street from being identified as obstacles and receiving assistance alerts to the driver at a station with the train doors open, at a station where a passenger crossing is located immediately before the train 100.
[0051] In the first and second embodiments, examples were described in which the judgment criteria employed either continuous detection threshold information, which indicates how many consecutive detections are required in the time-series data of the object detection results to determine a detection, or detection method information, which indicates whether the object detection results are determined as a detection by logical sum or logical product. The judgment criteria determined by the judgment criteria determination unit 122 may be other than either the continuous detection threshold information or the detection method information. For example, both the continuous detection threshold information and the detection method information may be used as judgment criteria. By setting the continuous detection threshold to a large value and using a logical product as the detection method, it is possible to make an object more difficult to detect. This makes it possible to further reduce false detections.
[0052] According to the embodiment described above, in the obstacle detection assistance system 120 installed in a rail transportation system, it is possible to reduce false alarms without interrupting driving assistance to the driver.
[0053] The above-described embodiment of the present invention can be summarized as follows.
[0054] (1) The obstacle detection assistance system 120 is a system that assists in obstacle detection during vehicle driving, and includes a driving assistance display unit 124 that displays driving assistance information that assists vehicle driving, an obstacle detection unit 123 that detects obstacles that hinder the driving of the vehicle, a judgment criterion determination unit 122 that determines judgment criteria used by the obstacle detection unit 123 to detect an obstacle, and an external environment recognition unit 121 that recognizes the external environment of the vehicle, wherein the judgment criterion determination unit 122 sets judgment criteria according to the external environment, and the judgment criterion is set to a threshold value for the number of times an obstacle is detected by the obstacle detection unit 123, or a degree of agreement between the detection results by the obstacle detection unit 123, or both the threshold value for the number of times an obstacle is detected and the degree of agreement between the detection results. As a result, the obstacle detection assistance system 120 can reduce false alarms without interrupting driving assistance to the driver.
[0055] (2) The judgment criteria determination unit 122 includes, as the external environment when changing the judgment criteria, one or more of the following: route environments such as stations, railroad crossings, and railway bridges; route shape environments such as straight lines and curves; time environments such as day and night; and train environments such as vehicle speed and vehicle position.
[0056] (3) When the determination criterion setting unit 122 determines that the external environment is one in which false detection is likely to occur, it changes the threshold value for the number of times an obstacle is detected to a larger value.
[0057] (4) When the determination criterion determination unit 122 determines that the external environment is one in which false detection is likely to occur, it changes the method for detecting the degree of coincidence of the detection results from logical sum to logical product.
[0058] (5) When the judgment criterion determination unit 122 determines that the external environment is one in which the vehicle is stopped and the door is open, it sets the threshold for the number of obstacle detections to a value equal to or greater than the maximum number of past detection result data that the system can hold.
[0059] (6) When the judgment criterion determination unit 122 determines that the external environment is one in which the vehicle is stopped and the doors are closed, it sets the threshold for the number of obstacle detections to a value less than the maximum number of past detection result data that the system can hold.
[0060] (7) The obstacle detection unit 123 determines only detected objects that exist within the travel area and within a distance at which the vehicle can stop as obstacle detection targets.
[0061] Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the gist of the present invention.
[0062] In the above figures, the control lines and information lines shown are those that are considered necessary for explanation, and do not necessarily show all the control lines and information lines that are actually implemented. For example, it can be considered that almost all components are actually connected to each other.
[0063] The above-described layout of the functional units of the obstacle detection assistance system 120 is merely an example. The layout of the functional units can be changed to an optimal layout in terms of the performance, processing efficiency, communication efficiency, etc. of the hardware and software included in the obstacle detection assistance system 120.
[0064] 100 Train 110 External sensor 120 Obstacle detection support system 121 External environment recognition unit 122 Criterion determination unit 123 Obstacle detection unit 124 Driving support display unit
Claims
1. An obstacle detection assistance system that assists in obstacle detection while a vehicle is traveling, comprising: a driving assistance display unit that displays driving assistance information, which is information that assists vehicle driving; an obstacle detection unit that detects obstacles that hinder the traveling of the vehicle; a judgment criterion determination unit that determines the judgment criteria used when the obstacle detection unit detects an obstacle; and an external environment recognition unit that recognizes the external environment of the vehicle, wherein the judgment criterion determination unit sets judgment criteria according to the external environment, and the judgment criterion is set to a threshold value for the number of times an obstacle is detected by the obstacle detection unit, or a degree of agreement of the detection results by the obstacle detection unit, or both the threshold value for the number of times an obstacle is detected and the degree of agreement of the detection results.
2. An obstacle detection support system as described in claim 1, characterized in that the judgment criteria determination unit includes, as the external environment when changing the judgment criteria, one or more of the following: route environment such as stations, crossings, and railway bridges; route shape environment such as straight lines and curves; time environment such as day or night; and train environment such as vehicle speed and self-position.
3. An obstacle detection support system as described in claim 1, characterized in that the judgment criterion determination unit changes the threshold for the number of obstacle detections to a larger value when it determines that the external environment is one in which false detections are likely to occur.
4. An obstacle detection support system as described in claim 1, characterized in that when the judgment criterion determination unit judges that the external environment is one in which erroneous detection is likely to occur, the method of detecting the degree of coincidence of the detection results is changed from logical sum to logical product.
5. An obstacle detection support system as described in claim 1, characterized in that when the judgment criterion determination unit judges that the external environment is that the vehicle is stopped and the door is open, it sets the threshold for the number of obstacle detections to a value equal to or greater than the maximum number of past detection result data that the system can hold.
6. An obstacle detection support system as described in claim 1, characterized in that when the judgment criterion determination unit judges that the external environment is that the vehicle is stopped and the doors are closed, it sets the threshold for the number of obstacle detections to less than the maximum number of past detection result data that the system can hold.
7. An obstacle detection support system as described in claim 1, characterized in that the obstacle detection unit only determines whether an obstacle is detected if the obstacle is present within the driving area and within a distance at which the vehicle can stop.
8. An obstacle detection support method for supporting obstacle detection during vehicle driving, comprising: a computer having at least a processor and a storage device displays driving support information that supports vehicle driving, detects obstacles that impede the driving of the vehicle, determines a judgment criterion for detecting the obstacle, recognizes an external environment of the vehicle, and when determining the judgment criterion, sets a judgment criterion according to the external environment, and sets as the judgment criterion a threshold value for the number of times the obstacle is detected when detecting the obstacle, or a degree of agreement of the detection results when detecting the obstacle, or both the threshold value for the number of times the obstacle is detected and the degree of agreement of the detection results.
Citation Information
Patent Citations
Obstacle detection support system
JP2022135627A
Train control system
JP2022123954A
Obstacle detection device and obstacle detection method
WO2019155569A1
Obstacle detection apparatus
WO2023032358A1
